Analüüsi eesmärk on:
* pideva muutuja averageGrade ennustamine.
* andmestiku struktuuri kirjeldamine (dimensionaalsuse vähendamine, klasterdamine).Algandmed on hoiustatud json failis, mille loen mällu ja kirjutan kolme erinevasse .csv faili, kus igaühes neist on esitatud kõik andmestiku omadused aga veidi erineval kujul, et seda hiljem lihtsam kasutada oleks.
import json
import csv
def convert_dict_to_csv(dict):
csv_file = "students"
csv_columns = ['id', 'uniid', 'lastTested', 'totalCommits', 'totalTestsRan', 'totalTestsPassed',
'totalDiagnosticErrors', 'differentSlugs', 'differentCourses', 'commitsStyleOK', 'averageGrade',
'medianGrade'] # Last one is the one being predicted
csv_columns_percent = ['totalDiagnosticErrors_per_totalCommits', 'totalTestsPassed_per_totalTestsRan',
'commitsStyleOK_per_totalCommits', ]
to_drop = ['id', 'uniid', 'lastTested', 'medianGrade']
to_drop_extra_only = ['totalCommits', 'totalTestsRan', 'differentCourses', 'totalTestsPassed',
'totalDiagnosticErrors', 'commitsStyleOK']
try:
for i in range(3):
with open(csv_file + ("_extra.csv" if i == 1 else ".csv" if not i else "_extra_only.csv"), 'w') as csvfile:
if i == 1:
csv_columns = csv_columns_percent + csv_columns
if i == 2:
csv_columns = [x for x in csv_columns if x not in to_drop_extra_only]
csv_columns = [x for x in csv_columns if x not in to_drop]
writer = csv.DictWriter(csvfile, delimiter=',', lineterminator='\n', fieldnames=csv_columns)
csvfile.write(
f"{len(dict)},{len(csv_columns) - 1}\n") # -1 because the last field is the one being predicted
writer.writeheader()
for data in reversed(dict):
if not i:
for drop in to_drop:
del data[drop]
if i:
for extra in csv_columns_percent:
left, right = extra.split("_per_")
if not data[right]:
data[extra] = 0
else:
data[extra] = data[left] / data[right]
if i == 2:
for drop in to_drop_extra_only:
del data[drop]
writer.writerow(data)
except IOError:
print("I/O error")
with open('example.json') as f:
dict_data = json.load(f)
convert_dict_to_csv(dict_data)
student.csv description:
Author: Enrico Vompa Date: 2020.03.20
Data is collected from Arete. (TalTech's automated testing service)
Dataset contains no NULL fields.
Nr of lines: 138 Nr of attributes: 8
Headers:
- totalCommits (integer) - Total number of commits from the student
- totalTestsRan (integer) - Total number of tests ran for the student
- totalTestsPassed (integer) - Total number of tests passed from the tests ran
- totalDiagnosticErrors (integer) - Total number of diagnostic errors. Diagnostic error = style error in code. Eslint for example.
- differentSlugs (integer) - Number of different exercises the given student has attempted
- differentCourses (integer) - Number of different courses the given student has attempted
- commitsStyleOK (integer) - Number of test jobs ran with no diagnostic errors
- averageGrade (float) - students' average score
students_extra.csv description:
Author: Enrico Vompa Date: 2020.04.11
Data is collected from Arete. (TalTech's automated testing service)
Dataset contains no NULL fields.
Nr of lines: 138 Nr of attributes: 11
Headers:
- totalDiagnosticErrors_per_totalCommits (float) - totalDiagnosticErrors divided by totalCommits
- totalTestsPassed_per_totalTestsRan (float) - totalTestsPassed divided by totalTestsRan
- commitsStyleOK_per_totalCommits (float) - commitsStyleOK divided by totalCommits
- totalCommits (integer) - Total number of commits from the student
- totalTestsRan (integer) - Total number of tests ran for the student
- totalTestsPassed (integer) - Total number of tests passed from the tests ran
- totalDiagnosticErrors (integer) - Total number of diagnostic errors. Diagnostic error = style error in code. Eslint for example.
- differentSlugs (integer) - Number of different exercises the given student has attempted
- differentCourses (integer) - Number of different courses the given student has attempted
- commitsStyleOK (integer) - Number of test jobs ran with no diagnostic errors
- averageGrade (float) - students' average score
students_extra_only.csv description:
Author: Enrico Vompa Date: 2020.04.11
Data is collected from Arete. (TalTech's automated testing service)
Dataset contains no NULL fields.
Nr of lines: 138 Nr of attributes: 5
Headers:
- totalDiagnosticErrors_per_totalCommits (float) - totalDiagnosticErrors divided by totalCommits
- totalTestsPassed_per_totalTestsRan (float) - totalTestsPassed divided by totalTestsRan
- commitsStyleOK_per_totalCommits (float) - commitsStyleOK divided by totalCommits
- differentSlugs (integer) - Number of different exercises the given student has attempted
- averageGrade (float) - students' average score
import pandas as pd
import numpy as np
import seaborn as sns
import matplotlib.pyplot as plt
import pandas as pd
from pandas.plotting import scatter_matrix
from sklearn.cluster import KMeans
def load_dataset(filename):
with open(filename) as csv_file:
data_file = csv.reader(csv_file)
temp = next(data_file)
n_samples = int(temp[0])
n_features = int(temp[1])
temp = next(data_file)
columns2= [x.strip() for x in temp]
data = np.empty((n_samples, n_features), dtype=np.int64)
target = np.empty((n_samples,), dtype=np.int64)
for i, sample in enumerate(data_file):
data[i] = np.asarray([int(round(float(x))) for x in sample[:-1]], dtype=np.int64)
target[i] = np.asarray(int(round(float(sample[-1]) / 10)) * 10, dtype=np.int64)
return Bunch(data=data, target=target),columns2
def getColumns(filename):
with open(filename) as f:
f.readline() # holder for dimensions
columns = [x.strip() for x in f.readline().split(",")]
return columns
Alustab andmete lugemisega
used_columns = getColumns("students.csv")
student_df = pd.read_csv("students.csv",
header=2,
names=used_columns)
student_df
print("Mean of __")
_ = [print(f" {x} is: {student_df[x].mean()}") for x in used_columns]
print()
print("Median of __")
_ = [print(f" {x} is: {student_df[x].median()}") for x in used_columns]
print()
print("Standard deviation of __")
_ = [print(f" {x} is: {student_df[x].std()}") for x in used_columns]
Eelnevatest Järeldab, et andmeid on vaja standartiseerida enne nende kasutust.
Histogrammid, et vaadata andmete jaotust
_ = [student_df.hist(column=x) for x in used_columns] ## Bigger pictures this way
Pairplots, et vaadata andmete jaotust ning uurida, millised kolumni väärtused on omavahel otseselt seotud graafiliselt
_ = scatter_matrix(student_df, figsize = (25, 25), alpha=0.8)
Sai joonistatud scatter_matix, et proovida leida seoseid testimiselt tulnud andmete ja keskmise hinde vahel. Aga suuri otseseid seoseid ei paista olevat. Selleks tuleks joonistada korrelatsioonimaatriks, et kindel olla.
Korrelatsioonimaatriks, et uurida millised kolumni väärtused on omavahel otseselt seotud
def histogram_intersection(a, b):
v = np.minimum(a, b).sum().round(decimals=1)
return v
student_df.corr(method=histogram_intersection)
Siin tunduvad suvalised numbrid olevat, niiet ka see väga palju midagi ei ütle.
Paneb graafikud samasse teljestikku, et vaadata, millised andmed on umbkaudu samas suurusjärgus
_ = student_df.plot(figsize=(20, 20))
Andmeid on igast suurusjärke, et tuleb standartiseerida kuidagi vastuseid
Joonistab korrelatsiooni koefitsendi maatriksi vahemikus -1 kuni 1, et oleks lihtsamini loetav ning visualiseerib need clustermap'ina ning heatmap'ina
def draw_coefficient_matrix(student_df):
X = student_df
y = student_df.columns
cm = np.corrcoef(X.T)
sns.set(font_scale=1.5)
hm = sns.heatmap(cm, cbar=True, annot=True, square=True, fmt='.2f',
annot_kws={'size': 6}, yticklabels=X.columns,
xticklabels=X.columns,
cmap=sns.diverging_palette(10, 220, sep=80))
plt.show()
km = sns.clustermap(cm, cbar=True, annot=True,fmt='.2f',
annot_kws={'size': 8}, yticklabels=X.columns,
xticklabels=X.columns,
cmap=sns.diverging_palette(10, 220, sep=80))
plt.show()
draw_coefficient_matrix(student_df)
Eelnevatest andmetest saab järeldata järgmist - Huvitaval kombel pole keskmise hinne otseselt seotud muude kolumni väärtustega. (Koefitsendi väärtused korrelatsioonimaatriksis real averageGrade on alla .4)
Erinevate kursuste arv on vastupidises suhtes erinevate ülesannete arvuga - Mille põhjendus on ilmselt see, et andmeid ei ole piisavalt pika aja peale. Peaks seda faili jooksutama uuesti kunagi tulevikus.
Ootuspäraselt on totalCommits, totalTestsRan ja totalTestsPassed omavahel samasuunalises suhtes.
Panen paar täiendavat fieldi sisse, millest saaks ehk midagi järeldada
student_df_extra = pd.read_csv("students_extra.csv",
header=2,
names=getColumns("students_extra.csv"))
draw_coefficient_matrix(student_df_extra)
Ning ootuspäraselt tekkis seos keskmise hinde ning totalTestsPassed_per_totalTestsRan vahel
Aga huvitaval kombel ei ole commitsStyleOK_per_totalCommits ega totalDiagnosticErrors_per_totalCommits vead kuidagi seotud keskmise hindega
from sklearn.metrics import silhouette_samples, silhouette_score
import matplotlib.cm as cm
def draw_silhouette_plots(X, range_n_clusters = range(2, 10)):
# See kood on võetud suuresti
# https://scikit-learn.org/stable/auto_examples/cluster/plot_kmeans_silhouette_analysis.html#sphx-glr-auto-examples-cluster-plot-kmeans-silhouette-analysis-py
distortions = []
for n_clusters in range_n_clusters:
# Create a subplot with 1 row and 2 columns
fig, (ax1, ax2) = plt.subplots(1, 2)
fig.set_size_inches(18, 7)
# The 1st subplot is the silhouette plot
# The silhouette coefficient can range from -1, 1 but in this example all
# lie within [-0.1, 1]
ax1.set_xlim([-0.1, 1])
# The (n_clusters+1)*10 is for inserting blank space between silhouette
# plots of individual clusters, to demarcate them clearly.
ax1.set_ylim([0, len(X) + (n_clusters + 1) * 10])
# Initialize the clusterer with n_clusters value and a random generator
# seed of 10 for reproducibility.
clusterer = KMeans(n_clusters=n_clusters, random_state=10)
cluster_labels = clusterer.fit_predict(X)
distortions.append(clusterer.inertia_)
# The silhouette_score gives the average value for all the samples.
# This gives a perspective into the density and separation of the formed
# clusters
silhouette_avg = silhouette_score(X, cluster_labels)
print("For n_clusters =", n_clusters,
"The average silhouette_score is :", silhouette_avg)
# Compute the silhouette scores for each sample
sample_silhouette_values = silhouette_samples(X, cluster_labels)
y_lower = 10
for i in range(n_clusters):
# Aggregate the silhouette scores for samples belonging to
# cluster i, and sort them
ith_cluster_silhouette_values = \
sample_silhouette_values[cluster_labels == i]
ith_cluster_silhouette_values.sort()
size_cluster_i = ith_cluster_silhouette_values.shape[0]
y_upper = y_lower + size_cluster_i
color = cm.nipy_spectral(float(i) / n_clusters)
ax1.fill_betweenx(np.arange(y_lower, y_upper),
0, ith_cluster_silhouette_values,
facecolor=color, edgecolor=color, alpha=0.7)
# Label the silhouette plots with their cluster numbers at the middle
ax1.text(-0.05, y_lower + 0.5 * size_cluster_i, str(i))
# Compute the new y_lower for next plot
y_lower = y_upper + 10 # 10 for the 0 samples
ax1.set_title("The silhouette plot for the various clusters.")
ax1.set_xlabel("The silhouette coefficient values")
ax1.set_ylabel("Cluster label")
# The vertical line for average silhouette score of all the values
ax1.axvline(x=silhouette_avg, color="red", linestyle="--")
ax1.set_yticks([]) # Clear the yaxis labels / ticks
ax1.set_xticks([-0.1, 0, 0.2, 0.4, 0.6, 0.8, 1])
# 2nd Plot showing the actual clusters formed
colors = cm.nipy_spectral(cluster_labels.astype(float) / n_clusters)
ax2.scatter(X[:, 0], X[:, 1], marker='.', s=30, lw=0, alpha=0.7,
c=colors, edgecolor='k')
# Labeling the clusters
centers = clusterer.cluster_centers_
# Draw white circles at cluster centers
ax2.scatter(centers[:, 0], centers[:, 1], marker='o',
c="white", alpha=1, s=200, edgecolor='k')
for i, c in enumerate(centers):
ax2.scatter(c[0], c[1], marker='$%d$' % i, alpha=1,
s=50, edgecolor='k')
ax2.set_title("The visualization of the clustered data.")
ax2.set_xlabel("Feature space for the 1st feature")
ax2.set_ylabel("Feature space for the 2nd feature")
plt.suptitle(("Silhouette analysis for KMeans clustering on sample data "
"with n_clusters = %d" % n_clusters),
fontsize=14, fontweight='bold')
plt.show()
plt.plot(range(2, 10), distortions)
plt.xlabel("Clusters")
plt.ylabel("Distortions")
plt.show()
Järgmiseks proovin klasterdamist, et vaadata, kas on võimalik andmeid klastritesse jaotada. Isiklikult arvan, et ei. Aga et kindel olla, siis proovin küünarnukki meetodit siluetidiagrammidel.
draw_silhouette_plots(student_df_extra.values)
Ootuspäraselt ei tekkinud ilusaid eristatavaid klastreid kui anda palju andmeid sisse. Järgmiseks proovin vaadata, mis juhtub, kui osad väljad ära võtta.
student_df_extra_only, extra_columns = load_dataset('students_extra_only.csv')
used_columns_extra = getColumns("students_extra_only.csv")
student_df_extra_only_pandas = pd.read_csv("students_extra_only.csv",
header=2,
names=used_columns_extra)
Siin on näha uued kasutusel olevad kolumnid ning nende jaotusdiagrammid, et saaks uurida, millised kolumnid kudidas omavahel seotud on.
_ = scatter_matrix(student_df_extra_only_pandas ,figsize = (25, 25), alpha=0.8)
draw_silhouette_plots(student_df_extra_only_pandas.values)
Vaadates nüüd klastrite kujunemist saab öelda, et erinevuseid leidub ning tudengeid saab klasterdada.
Klastrite arvu 3 puhul oli siluetti keskmine skoor: 0.7206387374319115, mis on päris hea.
Sellest saab järeldada, et tudengeid saab jagada kolme klastrisse.
Pideva väärtuse ennustamiseks loeb andmestiku uuesti sisse aga õigel kujul, et saaks neid lihtsamini kasutada
import matplotlib.pyplot as plt
import csv
from sklearn.preprocessing import StandardScaler, MinMaxScaler, PolynomialFeatures
from sklearn.decomposition import PCA
from sklearn.metrics import accuracy_score, confusion_matrix, f1_score
from sklearn.feature_selection import SelectKBest
from sklearn.model_selection import cross_val_score, StratifiedKFold, train_test_split, learning_curve, GridSearchCV
from sklearn.pipeline import Pipeline
from sklearn.tree import DecisionTreeClassifier, DecisionTreeRegressor
from sklearn.linear_model import Perceptron, LinearRegression, LogisticRegression
from sklearn.datasets.base import Bunch
from sklearn import utils
from sklearn.exceptions import ConvergenceWarning
from sklearn.ensemble import RandomForestRegressor
import warnings
warnings.filterwarnings("ignore", category=FutureWarning)
warnings.filterwarnings("ignore", category=Warning)
warnings.filterwarnings("ignore", category=ConvergenceWarning)
student_df2, columns2 = load_dataset('students.csv')
print("Average grades after conversion:")
X_train, X_test, y_train, y_test = train_test_split(student_df2.data, student_df2.target, test_size=0.25, random_state=42)
print("Train:\n", y_train)
print("Test:\n", y_test)
Järgmiseks vaatab, kas andmeid on piisavalt, hinnates selle õppimiskõverat
def validation_curve():
train_sizes, train_scores, test_scores =\
learning_curve(estimator=pipe_lr,
X=X_train,
y=y_train,
cv=10)
train_mean = np.mean(train_scores, axis=1)
train_std = np.std(train_scores, axis=1)
test_mean = np.mean(test_scores, axis=1)
test_std = np.std(test_scores, axis=1)
%matplotlib inline
plt.plot(train_sizes, train_mean,
color='blue', marker='o',
markersize=5, label='training accuracy')
plt.fill_between(train_sizes,
train_mean + train_std,
train_mean - train_std,
alpha=0.15, color='blue')
plt.plot(train_sizes, test_mean,
color='green', linestyle='--',
marker='s', markersize=5,
label='validation accuracy')
plt.fill_between(train_sizes,
test_mean + test_std,
test_mean - test_std,
alpha=0.15, color='green')
plt.grid()
plt.xlabel('Number of training samples')
plt.ylabel('Accuracy')
plt.legend(loc='lower right')
plt.ylim([0, 1.0])
plt.tight_layout()
plt.show()
def printScore(): # Cross validation
scores = cross_val_score(estimator=pipe_lr,
X=X_train,
y=y_train,
cv=10)
print('CV keskmine täpsus: %.3f' % abs(np.mean(scores)), "+/- %.3f" % np.std(scores))
y_pred = [int(round(x / 10) * 10) for x in pipe_lr.predict(X_test)]
print('F1 keskmine täpsus: %.3f' % f1_score(y_test, y_pred, average='micro'))
print('Pipeline keskmine täpsus: %.3f' % pipe_lr.score(X_test, y_test))
def print_confusion_matrix():
print("\nEksimaatriks, kus telgedel on ennustatavad (y-telg) ja ennustatud(x-telg) väärtused:")
pipe_lr.fit(X_train, y_train)
y_pred = [int(round(x / 10) * 10) for x in pipe_lr.predict(X_test)]
confmat = confusion_matrix(y_true=y_test, y_pred=y_pred)
cols = [str(i) for i in range(100 - (len(confmat) * 10) + 10, 110, 10)]
df_cm = pd.DataFrame(confmat, index=cols, columns=cols)
plt.figure(figsize = (10,7))
sns.heatmap(df_cm, annot=True)
def use_pipe_lr():
validation_curve()
print_confusion_matrix()
printScore()
pipe_lr = Pipeline([('tree', DecisionTreeRegressor())])
use_pipe_lr()
print("Vaadates jaotust on DecisionTreeRegressor paljutõotav")
pipe_lr = Pipeline([('sc1', SelectKBest(k=2)),
('tree', DecisionTreeClassifier(max_depth=1))])
use_pipe_lr()
print("Päris hea tulemus. Hiljem proovin selle peal hüperparameetrite tuunimist")
pipe_lr = Pipeline(steps = [('lr', LinearRegression())])
use_pipe_lr()
print("Pipeline'ist on asju puudu")
pipe_lr = Pipeline([('norm', StandardScaler()),
('lr', LinearRegression())])
use_pipe_lr()
print("Väärtused on kuidagi mööda ilma laiali")
pipe_lr = Pipeline([('norm', MinMaxScaler()),
('lr', LinearRegression())])
use_pipe_lr()
print("Ka MinMAxScaler'iga on väärtused igasse kanti")
pipe_lr = Pipeline(steps = [('pf', PolynomialFeatures()),
('lr', LinearRegression())])
use_pipe_lr()
print("See on väga vale")
Lineaarne regressioon ei tundu sobiva variandina hetkel
pipe_lr = Pipeline([('rfr', RandomForestRegressor())])
use_pipe_lr()
print("Siin ei tundu jaotus eriti hea")
pipe_lr = Pipeline([('sc1', SelectKBest(k=2)),
('rfr', RandomForestRegressor(n_estimators=50, criterion='mse'))])
use_pipe_lr()
print("RandomForestRegressor nii hea selle ülesane lahendamiseks pole")
pipe_lr = Pipeline([('sc', MinMaxScaler()),
('pca', PCA(n_components=2)),
('per', Perceptron())])
use_pipe_lr()
print("Siin kaotatakse kuidagi väikesed väärtused ära")
pipe_lr = Pipeline([('sc', StandardScaler()),
('pca', PCA(n_components=2)),
('per', Perceptron())])
use_pipe_lr()
print("Midagi hakkab looma. Hiljem katsetatakse hüperparameetritega")
pipe_lr = Pipeline(steps = [('lr', LogisticRegression(random_state=0, solver="liblinear"))])
use_pipe_lr()
print("See on vägagi paljutõotav")
pipe_lr = Pipeline(steps = [('pf', PolynomialFeatures()),
('lr', LogisticRegression())])
use_pipe_lr()
print("PolynomialFeatures väga ei sobi")
pipe_lr = Pipeline(steps = [('mms', MinMaxScaler()),
('lr', LogisticRegression())])
use_pipe_lr()
print("MinMaxScaler on juba parem")
pipe_lr = Pipeline(steps = [('ss', StandardScaler()),
('lr', LogisticRegression())])
use_pipe_lr()
print("StandardScaler sobib veidi paremini")
pipe_lr = Pipeline([('sc1', StandardScaler()),
('pca', PCA()),
('clf', LogisticRegression())])
use_pipe_lr()
print("PCA võib vahele visata, kuna worst case see ei tee midagi. Saadab selle konfiguratsiooni ka hüperparameetritega testimisele")
pipe_lr = Pipeline([('sc1', MinMaxScaler()),
('pca', PCA(n_components=2)),
('clf', LogisticRegression())])
use_pipe_lr()
print("StandardScaler tundus parem. Proovib tuunida hüperparameetritega")
Mudeleid sai hinnatud kolmel moel:
* CV
* F1
* Pipeline.score
Vaadates eksimaatrikseid ning võrrelda täpsuseid siis kõige täpsem neist oli cross validation, mida ka edasipidi kasutan
Tulemused, mida proovib hüperparameetritega parendada:
* SelectKBest(k=2) ning DecisionTreeClassifier(max_depth=1) andis tulemuseks 0.492 +/- 0.127
* StandardScaler(), PCA(n_components=2) ning Perceptron() andis tulemuseks 0.472 +/- 0.137
* StandardScaler(), PCA(n_components=2) ning LogisticRegression() andis tulemuseks 0.538 +/- 0.141DecisionTreeClassifier tuunimine hüperparameetritega
pipe_lr = Pipeline([('skb', SelectKBest()),
('tree', DecisionTreeClassifier())])
param_grid = [{'skb__k': [1, 2, 3, 4, 5],
'tree__splitter': ["best", "random"],
'tree__max_depth': [1, 2, 3, 4, 5, 10, 100],
'tree__criterion': ["gini", "entropy"]
}]
gs = GridSearchCV(estimator=pipe_lr,
param_grid=param_grid,
scoring='accuracy',
cv=5,
n_jobs=-1)
gs = gs.fit(X_train, y_train)
print('Täpsus: ', gs.best_score_)
print(gs.best_params_) # {'skb__k': 4, 'tree__criterion': 'entropy', 'tree__max_depth': 2, 'tree__splitter': 'best'}
pipe_lr = Pipeline([('skb', SelectKBest(k=4)),
('tree', DecisionTreeClassifier(criterion="entropy", max_depth=2, splitter="best"))])
use_pipe_lr()
pipe_lr = Pipeline([('sc', StandardScaler()),
('pca', PCA(n_components=2)),
('per', Perceptron())])
param_grid = [{'pca__n_components': [1, 2, 3, 4, 5],
'per__max_iter': [10, 100, 500, 1000],
'per__penalty': ["l2", "l1", "elasticnet", "none"]
}]
gs = GridSearchCV(estimator=pipe_lr,
param_grid=param_grid,
scoring='accuracy',
cv=10,
n_jobs=-1)
gs = gs.fit(X_train, y_train)
print('Täpsus: ', gs.best_score_)
print(gs.best_params_) # {'pca__n_components': 2, 'per__max_iter': 100, 'per__penalty': 'l1'}
pipe_lr = Pipeline([('sc', StandardScaler()),
('pca', PCA(n_components=2)),
('per', Perceptron(max_iter=100, penalty="l1"))])
use_pipe_lr()
pipe_lr = Pipeline([('sc1', StandardScaler()),
('pca', PCA()),
('lgr', LogisticRegression())])
param_grid = [{'pca__n_components': [1, 2, 3, 4, 5],
'lgr__max_iter': [10, 100],
'lgr__solver': ["newton-cg", "lbfgs", "sag", "saga"],
'lgr__penalty': ["l2", "none"]
}]
gs = GridSearchCV(estimator=pipe_lr,
param_grid=param_grid,
scoring='accuracy',
cv=10,
n_jobs=-1)
gs = gs.fit(X_train, y_train)
print('Täpsus: ', gs.best_score_)
print(gs.best_params_) # {'lgr__max_iter': 100, 'lgr__penalty': 'none', 'lgr__solver': 'lbfgs', 'pca__n_components': 5}
pipe_lr = Pipeline([('sc1', StandardScaler()),
('pca', PCA(n_components=5)),
('lgr', LogisticRegression(max_iter=10, penalty='l2', solver='newton-cg'))])
use_pipe_lr()
Parim mudel keskmise hinde arvutamiseks järgmine:
* StandardScaler(), PCA(n_components=5) ning LogisticRegression(max_iter=10, penalty='l2', solver='newton-cg')Kokkuvõte:
Tudengeid andis jagada kolme klastrisse nii, et siluetti skoor oli 0.72, mis oli parim skoor võimalike klastrite arvu poolest. Klastrid moodustusid tudengite õppeedukuse põhjal
Pideva muutuja ennustamisel valisin pidevaks muutujaks averageGrade'i. Parim mudel selle ennustamiseks oli pipeline, mis koosnes järgnevatest komponentidest:
- StandardScaler()
- PCA(n_components=5)
- LogisticRegression(max_iter=10, penalty='l2', solver='newton-cg')
Ning mis koondus ilusti ning andis tulemuseks 0.55 +/- 0.137, mis ole just parim tulemus. Aga Kui järgmine aasta seda faili uuesti jooksutada, siis peaks see tulemus tõusma, kuna siis on juba rohkem andmeid tudengite testimiste kohta.
Kokkuvõte:
Tudengeid andis jagada kolme klastrisse nii, et siluetti skoor oli 0.72, mis oli parim skoor võimalike klastrite arvu poolest. Klastrid moodustusid tudengite õppeedukuse põhjal
Pideva muutuja ennustamisel valisin pidevaks muutujaks averageGrade'i. Parim mudel selle ennustamiseks oli pipeline, mis koosnes järgnevatest komponentidest:
- StandardScaler()
- PCA(n_components=5)
- LogisticRegression(max_iter=10, penalty='l2', solver='newton-cg')
Ning mis koondus ilusti ning andis tulemuseks 0.55 +/- 0.137, mis ole just parim tulemus. Aga Kui järgmine aasta seda faili uuesti jooksutada, siis peaks see tulemus tõusma, kuna siis on juba rohkem andmeid tudengite testimiste kohta.
Kokkuvõte:
Tudengeid andis jagada kolme klastrisse nii, et siluetti skoor oli 0.72, mis oli parim skoor võimalike klastrite arvu poolest. Klastrid moodustusid tudengite õppeedukuse põhjal
Pideva muutuja ennustamisel valisin pidevaks muutujaks averageGrade'i. Parim mudel selle ennustamiseks oli pipeline, mis koosnes järgnevatest komponentidest:
- StandardScaler()
- PCA(n_components=5)
- LogisticRegression(max_iter=10, penalty='l2', solver='newton-cg')
Ning mis koondus ilusti ning andis tulemuseks 0.55 +/- 0.137, mis ole just parim tulemus. Aga Kui järgmine aasta seda faili uuesti jooksutada, siis peaks see tulemus tõusma, kuna siis on juba rohkem andmeid tudengite testimiste kohta.